Elphel image quality?

[Andrey Filippov]

Originally Posted by Juan M. M. Fiebelkorn

So, this is not suitable?

http://www.kodak.com/global/plugins/...uctSummary.pdf

it says it runs to 64 fps, 5.5 micron pixel.

Does it make any sense having a peltier for cooling the sensor box?

Peltier cooler only makes sense with slow-scan CCDs and for astronomic or similar applications (where most pixels are nearly black). Then it is important that the noise caused by the thermal electrons (accumulated during frame readout) will not add significantly to the pixels in interest. As you see from the datasheet the readout noise is 12e- (~1/2000 of the full scale), so the dark charge accumulated during the frame readout should be less than ~100e- so square root of it (that is a random part that you can not subtract) is less that the readout noise. As soon as the pixel charge caused by the light (useful) is more than the same 100e- level (0.5% of the saturation level) and more that the thermal current integrated during the frame readout time (and readout time is short at high fps) - the shot noise will become the major part of the overall noise. The thermal current decreases approximately twice for each 7C of the cooling, so for the same sensor going from 1 fps to 30fps (and some CCD are scanned much slower than 1 fps) gives the equivalent of log2(30)*7 ~= 35degrees C cooling.

I am considering that sensor, but only after the 373 will be ready and if there will be a good application for such project - it is different price range, complexity and size of the sensor board.

[Juan M. M. Fiebelkorn]

Seeing that many motherboards sporting USB 3.0 have been announced.Is there any possibility for this new camera design to use USB 3.0 ?

Would it be better than using the Gethernet option?

[Andrey Filippov]

There are multiple reasons why there will be no USB 3.0:
* I do not know any microcontroller that has it implemented, there are not many with GigE;
* There will be not enough data to use 5Gb/s (500 MB/s) bandwidth - TI DaVinci has just about 100MB/s from the FPGA to CPU; Of course it is possible to implement USB 3.0 in the FPGA, but it is _a lot_ of work that will be wasted - such dedicated chips will become available in a few years.
* It will not be possible to plug that camera into a network, only to connect it to some computer. So it will be not a network camera, but kind of a web-camera :-)
...
* And the last one - 373 PCB circuit and PCB design is mostly done - I'm now working on verifying the design and fixing the problems I can find. With such PCB density it will be extremely difficult to fix anything on a prototype when it will be built. Hope to release files to production by the next week.

Andrey

[Juan M. M. Fiebelkorn]

* PD720200 | USB3.0 Host Controller | Product Lineup | USB | Interface | Products | NEC Electronics
* You already have the GigE on the Davinci, it's just another option
*The last one is a good point!! :D

[Andrey Filippov]

Originally Posted by Juan M. M. Fiebelkorn

* PD720200 | USB3.0 Host Controller | Product Lineup | USB | Interface | Products | NEC Electronics

1 - It is not a part of the microcontroller (CPU), you need PCIe port to connect it. DaVinci has an option of just PCI - too slow for USB 3.0, and if you use it - you'll have to give up other peripherals - SATA port (through ATA on DaVinci), high speed connection to the FPGA.
2 - there is no datasheet there - just "product brief". That means that you have to sign NDA with them - do you think that will work for the open source camera?

Andrey

[Juan M. M. Fiebelkorn]

1-Didn't know that.I thought I read somewhere Spartan 6 supports PCIe and because you mentioned the FPGA for the USB3......Forget it :D

2-Well, about the product brief and NDAs, the same happens to me with Aptina sensors (no access to datasheets although I got an account.Asking for NDA), and because of that I thought in this situation it was not different.Sorry.

[Andrey Filippov]

Originally Posted by Juan M. M. Fiebelkorn

2-Well, about the product brief and NDAs, the same happens to me with Aptina sensors (no access to datasheets although I got an account.Asking for NDA), and because of that I thought in this situation it was not different.Sorry.

Yes, that is a problem. We do have datasheet for Aptina (it was Micron) 5MPix sensor (as well as for earlier 1.3-2-3MPix ones) that we got years ago without signing any NDA but that's all. Even if we'll ever sign NDA with Aptina we need written waiver that 2 pieces of information should be out of the NDA scope:

1 - device pinout that we need to include in our circuit diagrams that we provide;
2 - register addresses/descriptions that are used in the driver code.

It was much better to deal with Kodak - probably they are proud of their sensors and are not ashamed of releasing the full datasheets without any NDA. Same with Axis, Texas Instruments, Xilinx and many other component manufacturers that we use in Elphel products. Unfortunately Kodak does not have small fast high resolution CMOS sensors.

Andrey

[Juan M. M. Fiebelkorn]

Is it the same situation with other Image sensor manufacturers? Like Omnivision, Sony, Toshiba, etc....